Large Language Models are architecturally stateless systems where model weights are completely frozen during inference and do not update based on conversation content